58 results about "Risk category" patented technology

The invention relates to the technical field of computers, and discloses a model content security control method and system, electronic equipment and a storage medium, and the method comprises the steps: combining a received user input with a historical dialogue record of a current session to form a dialogue context; performing risk identification based on the dialogue context, generating at least one risk category and a corresponding initial risk value, and generating a risk integral corresponding to the current input according to a preset integral strategy; accumulating the risk points to accumulated risk points of the current session; comparing the accumulated risk integral with a dynamic risk threshold value obtained by calculation according to a reference threshold value, a logarithmic function attenuation item of the dialogue round and a user historical behavior adjustment item; and when the accumulated risk integral reaches or exceeds a dynamic risk threshold value, triggering a preset risk management and control measure. According to the method, the risk content can be effectively identified and controlled, attacks bypassing a security mechanism through slow induction can be defended, and the security boundary is not easy to detect.

The systems and methods disclosed herein receive alphanumeric characters defining operative boundaries for expected model use cases, along with operational data. The expected model use cases share common attributes, which are used by a first AI model to construct observed model use cases from the operational data. Each observed model use case includes features such as a text-based description, expected input and output,? AI model(s) generating the expected output from the input, and / or data supporting the AI models. For each observed model use case, a second AI model maps the alphanumeric characters and features to a risk category, selecting from multiple risk categories based on the level of risk associated with the features. The system identifies criteria for the observed model use case within the alphanumeric characters and generates gaps by comparing the criteria with the features of the observed model use case.

The invention relates to the technical field of artificial intelligence, and discloses a feature screening method and device based on artificial intelligence, equipment and a storage medium, and the method comprises the steps: obtaining a feature factor which affects the pricing of vehicle insurance; preprocessing the feature factors to obtain features influencing car insurance pricing and risk category labels of the features; through a support vector machine and an arrangement importance algorithm, according to the features and the risk category labels of the features, candidate features are screened out from the features, and the importance degrees of the candidate features are obtained; and through a quantum annealing algorithm, according to the risk category labels and the importance degrees of the candidate features, target features influencing vehicle insurance pricing are screened out from the candidate features. The method can be applied to feature screening in a financial vehicle insurance pricing scene, the efficiency and quality of feature screening can be remarkably improved, and the feature screening difficulty is effectively reduced.

According to an embodiment, disclosed is a system comprising a processor configured to define, one or more risk categories for monitoring a risk associated with a clinical trial, wherein the risk categories comprise one or more risk elements; calculate, a first risk profile data of the risk categories based on a risk factor and a weighting assigned to the risk elements; generate, a machine learning (ML) model; train, the ML model; receive, a second risk profile data; analyse, the second risk profile data to identify a pattern based on the first risk profile data using a database; predict, an overall risk score; recommend, one or more of a type of monitoring, a level of monitoring, and the overall risk score; and wherein the ML model comprises a feed-back layer to enable continuous learning and improve the prediction of the overall risk score and monitoring decisions of the clinical trial.

This application provides a data processing method, apparatus, device, and medium. The method includes: acquiring a first text containing business text data; performing a risk assessment on the first text to obtain a risk category result corresponding to the first text; if the risk category result is a first risk category, acquiring keywords of the business text data and searching for the keywords in a standard database; if a request text matching the keywords is found in the standard database, determining the feedback text corresponding to the request text as the business processing result corresponding to the business text data; if no request text matching the keywords is found in the standard database, performing text search processing on the business text data in a target knowledge graph to obtain a business processing result matching the business text data. Implementing this application embodiment can improve the security of text data.

The invention relates to the field of risk classification prediction, and provides a medical institution risk classification prediction method and device based on an attention mechanism, and the method comprises the steps: carrying out the analysis of a to-be-classified file through a medical risk classification prediction model, and obtaining a feature map and different risk categories; calculating the weight of each pixel point in the feature map through a medical risk classification prediction model, and adjusting the weight of each pixel point according to the calculated weight to obtain a weighted feature map; calculating the weight of each channel in the feature map through a medical risk classification prediction model, and adjusting the weight of each channel according to the calculated weight to obtain an adjusted feature map; features are extracted based on an attention mechanism, the probabilities of different risk categories are calculated through a normalization function, and the risk category corresponding to the maximum probability value serves as a prediction result to be returned; accurate identification and multi-modal identification of medical institution terminologies are realized, and the accuracy and automatic processing of risk classification and identification are improved.

This invention discloses a method for quantitatively evaluating the edge properties of complex scenarios based on a multi-category risk field model, belonging to the field of autonomous driving technology. The method includes: analyzing and classifying temporal scene elements based on the PEGASUS system; constructing a large-scale risk field for unidentified scene elements; constructing a medium-scale risk field for low-confidence identified scene elements; constructing an anisotropic driving behavior risk field for high-confidence identified scene elements; superimposing multiple risk fields within the interaction area; and quantifying the edge properties of the scenario using the minimum travel cost of the main vehicle's path. This invention achieves accurate quantification of the edge properties of complex scenarios by differentially modeling multiple risk sources and superimposing field strengths, utilizing the optimal path cost. It solves the problem that existing methods struggle to uniformly characterize multiple risk categories and overall travel costs, providing strong support for risk assessment and edge scenario mining for autonomous vehicles.

The invention discloses an SQL statement intelligent auditing and risk early warning method for financial services, and relates to the technical field of financial service intelligent auditing, and the method comprises the following steps: 1, defining a financial SQL core risk type and a quantitative auditing dimension; 2, a full-dimensional sensing module is constructed, multi-source data related to the SQL to be audited are collected, and the multi-source data at least comprise SQL statement data, database operation environment data, financial service rule data, supervision compliance data and historical SQL execution log data. According to the method, financial SQL auditing scene elements are mapped into four types of element individuals of SQL, risk, compliance and business, association strength among the individuals is quantified in combination with a connection weight algorithm, a virtual social network is constructed, four types of core risks of multi-risk-dimension collaborative judgment, sensitive data coverage, compliance authorization, business logic and performance overload are realized, and the risk level of the financial SQL auditing scene is improved. And the one-sidedness of a traditional static rule is avoided.

The invention belongs to the technical field of artificial intelligence and natural language processing, and provides a customer service quality inspection system based on large language model distillation and a deployment method thereof. The method comprises the steps that multiple groups of first annotation data are obtained, and the data are generated through risk annotation of multiple groups of historical dialogues between customer service staff and customers and comprise dialogue information and corresponding risk category annotations; inputting each group of first annotation data into a trained first semantic model, and outputting a first probability label (namely a first quality inspection result) containing semantic judgment logic and different quality inspection category probability distribution information; and in combination with the plurality of first probability tags and the first annotation data, training to obtain a second semantic model of which the parameter scale is significantly smaller than that of the first semantic model. According to the method, the quality inspection efficiency can be improved while the quality inspection effect is guaranteed, and the hardware resource cost of model deployment is reduced.

A method for threat detection may include obtaining data that is related to a series of digital activities performed with accounts on a channel through which employees of an enterprise can communicate with other employees of the enterprise or accounts external to the enterprise. The method may include parsing the data to identify an attribute of each digital activity. The method may include generating one or more metrics indicative of a threat posed by a respective digital activity of the series of digital activities. The method may include generating a plurality of digital profiles for at least some of the employees of the enterprise based on the series of digital activities and comprising the one or more metrics. The method may include generating a graphical user interface indicating a risk category associated with at least some of the digital activities of the series of digital activities.

The invention discloses a transaction risk identification method and device, electronic equipment and a storage medium, and relates to the technical field of financial science and technology, the method comprises the following steps: collecting structured transaction data and unstructured multi-modal data in a bank transaction process, the unstructured multi-modal data comprising a voice signal, a video signal and text data; performing feature extraction and coding processing on the unstructured multi-modal data, and mapping voice features, video features and text features to a unified semantic space; based on a multi-modal fusion model, performing fusion modeling on the mapped multi-modal features through a cross-modal interaction mechanism, and generating a transaction risk score and a risk category; and executing a corresponding real-time risk intervention strategy according to the risk score and the risk category. According to the method and the device, the structured transaction data and the unstructured multi-modal data in the bank transaction process are fused through the multi-modal fusion model, so that the risk identification dimension is improved, and the change of a fraud strategy can be quickly adapted.

The application discloses a kind of server security access method, device, electronic equipment and readable storage medium, applied to server, server is connected with one or more client communication, method includes: according to server log file, the access log of each user corresponding to each client is determined;Each access log is analyzed using risk assessment rule, and the target risk category that corresponding user hits in risk assessment rule is determined;Risk assessment rule includes the combination of one or more of API connection failure assessment rule, server resource access frequency assessment rule, off-site login assessment rule, high concurrent access assessment rule;According to the corresponding relationship between risk category and restriction rule, the target restriction rule corresponding to target risk category is determined;The user that hits corresponding target risk category is limited using target restriction rule.This scheme can be more accurate user risk assessment to client user and realize more flexible risk user control.

The application relates to a senile encephalopathy risk prediction method and system based on an attention mechanism and feature weighting optimization, which comprises the following steps: S1, multi-source heterogeneous data acquisition and preprocessing; S2, calculating the attention weight of each feature on risk prediction based on a multi-head self-attention mechanism, and constructing a dynamic feature weighting representation; S3, introducing a Mish activation function to perform nonlinear transformation on the weighted features, and enhancing the fitting capability of the model to complex nonlinear relationships; S4, constructing a feature selection mechanism based on the attention weight, and adaptively screening a key feature subset; S5, based on the key feature subset, adopting a Transformer encoder to construct a senile encephalopathy risk prediction model, and outputting the probability and risk grade of each risk category. The application can realize more accurate, efficient and adaptive senile encephalopathy risk prediction.

The application discloses a generative AI content security automatic testing method, relates to the technical field of artificial intelligence security, and comprises the following steps: acquiring a batch of test prompt templates, constructing a multi-dimensional risk classification label system to automatically label final content data, generating content risk categories and severity levels, combining decision interception logs to calculate content risk security scores, deploying state monitoring points in the internal network layer of the model, calculating internal state risk scores, accurately evaluating the risks of generated content, calculating comprehensive reward values through a pre-defined reinforcement learning reward function based on the content risk security scores and the internal state risk scores, optimizing and updating adjustable parameters of a prompt word generator by using a policy gradient algorithm, calculating security boundary ambiguity indexes under different semantic attack vectors according to accumulated data in multiple test cycles, generating targeted model optimization suggestions, and driving target generative AI model security policy incremental updates.

Methods and server systems for categorizing payment cards for fraud prevention are described herein. Method performed by a server system includes accessing a card candidate set including relevant payment card(s), each payment card of the relevant payment card(s) being associated with multiple features. Method includes segregating the features into a set of risk-related features and a set of transactional features. Method further includes generating, by Machine Learning (ML) model(s), a riskiness score for each payment card based on the set of risk-related features. Method includes performing for each payment card: assigning a risk category based on the riskiness score and risk categorization criteria, and assigning a transactional category based on the set of transactional features and transaction behavior criteria. Method includes generating a recommendation message for an issuer based on the risk category and the transactional category.

The invention relates to a massive social session risk clue mining method and device based on a large language model, a storage medium and computer equipment, and the method comprises the steps: generating a first cue word based on massive social sessions and a first cue word template, and inputting the first cue word into the large language model, obtaining multiple pieces of risk information output by the large language model and the risk category of each piece of risk information; generating a second cue word based on the multiple pieces of risk information, the risk category of each piece of risk information and a second cue word template, and inputting the second cue word into the large language model to obtain similar risk merging content output by the large language model; and generating a third cue word based on the similar risk merging content and the third cue word template, and inputting the third cue word into the large language model to obtain a risk clue mining report output by the large language model. According to the method, the technical problems of low efficiency, insufficient accuracy, large manual workload, high concealment, difficulty in clue recognition and the like in mass social communication data analysis can be solved.

Embodiments of the present specification provide a method and device for training a risk control model and predicting a risk category. The method for training the risk control model comprises: obtaining each first training sample, the first training sample comprising a first feature value corresponding to an attribute feature of a business object and a category label indicating whether the business object has a business risk; determining a second feature value corresponding to a scenario feature, adding the second feature value to the corresponding first training sample to form a second training sample; constructing a first decision tree through node splitting based on each second training sample, the process of splitting for a current node comprising: splitting according to a splitting purity of any splitting condition in a plurality of candidate splitting conditions of the current node; regarding the scenario feature as a category feature during the splitting process; and determining a risk control model for classifying business objects based on the first decision tree. The complexity of the model system can be simplified, and the model has better performance.

The invention relates to a computer-implemented method for processing data relating to changes in cloud services (1), comprising: retrieving data relating to a change in a cloud service (1); reducing the data relating to the change; comparing the reduced data with stored data relating to already detected changes; comparing whether the change relates to an active service if both conditions are met; processing the reduced data in a risk assessment process, comprising: identifying risk categories; a risk analysis for textually specifying risk factors to the risk categories; a risk assessment; proposing measures, wherein these sub-steps are carried out by a user supported by software and in advance by a large language model, wherein the result of the process processed by the large language model is provided to the user in addition to the data relating to the new change.

The invention relates to the technical field of resource management, and provides a supply chain risk quantification method based on large language model weak supervised learning. The method comprises the following steps: obtaining a multi-time-period operation disclosure text of a target enterprise and cleaning clauses to form a structured statement set; screening a potential risk statement subset based on the supply chain risk keyword library; calling a large language model to carry out weak supervision labeling to generate a risk category label; training a small risk sentiment classification model by using a label, reasoning a full amount of statements, and outputting a probability value of each risk category; aggregating the probability according to a time period to obtain a semantic distribution vector to represent a risk semantic state; and constructing a semantic residual image for the adjacent periodic vector difference values, and calculating a supply chain risk trend index according to the semantic residual image to realize risk evolution trend quantification.

Systems and methods for reducing false positives for financial transaction fraud monitoring using machine learning techniques. Using an original model for separating transactions into high risk and low risk categories for fraud, transactions falling into the high-risk category may be labeled as false positive or true positive. The labels and data associated with the transactions may be used to train two or more sequential false positive reduction models (FPRMs). Once a desired minimal amount of false positive labels are found in transactions labeled high risk by one of the FPRMs, that one of the FPRMs becomes the trained FPRM. Next, an additional transaction is processed using the original model and the trained FPRM, which then determines whether the additional transaction is at a high risk of being fraudulent.

The invention discloses a system risk monitoring method and device, computer equipment and a medium, and the method comprises the steps: building a ship equipment development full-period risk feature database, collecting stock ship project data, defining the types and exclusive features of six first-level risks and subordinate second-level risks, and carrying out the collection of the stock ship project data; setting a scoring standard, mapping stock project risk features into vectors, and performing associated storage; extracting information of a to-be-reviewed project, calculating dynamic risk indexes of each secondary risk category of the to-be-reviewed ship project by using the dynamic risk index model, and dividing risk levels; determining a target risk feature vector according to the ship type and the development stage of the to-be-reviewed ship project; calculating the cosine similarity between the target risk feature vector and all stock ship project risk feature vectors in a ship equipment development full-cycle risk feature database, clustering to obtain a similar project group, and counting risk data to generate a prompt statement; and establishing a risk coping scheme library, selecting an initial scheme and setting an execution plan.

The invention provides a customer risk identification method which can be applied to the technical field of artificial intelligence, and relates to the application of a large model in the fields of information security and financial science and technology. The customer risk identification method comprises the steps of processing customer business data according to a preset classification rule, and generating a to-be-evaluated index library; comparing the to-be-evaluated index library with the industry reference data to obtain index thresholds in one-to-one correspondence with to-be-evaluated indexes; performing risk assessment on the to-be-assessed index based on the index threshold, and generating a risk report, the risk report including a risk list and a risk overall evaluation generated based on the risk index, and a check list generated based on the to-be-assessed index not the risk index; and carrying out risk category identification on the risk indexes in the risk list, and adding the identified risk category to a risk report. The invention further provides a customer risk identification device and equipment, a storage medium and a program product.

The embodiment of the invention discloses a video detection method and device. According to the method, firstly, a to-be-detected video is obtained, an optical flow field between adjacent frames in the video is calculated, then motion components (such as vertical or horizontal displacement) related to a preset risk category are extracted, an initial risk detection result of each frame is generated, then statistics is carried out on all the initial detection results to form structured risk statistical information, and the risk statistical information is obtained. And filling the information into a prompt template corresponding to a preset risk category, automatically generating a risk detection prompt statement, finally inputting the statement into a visual large model, and outputting a final risk detection result. The method does not need to depend on complex modules such as target detection, key dynamic features can be captured and initial detection can be completed only through optical flow, then risk detection is achieved through combination of the prompt template and the visual large model, the calculation overhead can be reduced, the risk identification efficiency and semantic accuracy can be improved, meanwhile, the good generalization ability is achieved, and the method is suitable for popularization and application. The method can be adapted to various risk scenes and has expansibility.

Method ( 100 ) implemented through a distributed computer system (2 ) to dynamically determine at least one travel or stay risk index in at least one geographical area, comprising the steps of : - defining ( 101 ) a plurality of risk categories, each risk category corresponding to a type of risk to which a person may be exposed when travelling or staying in said geographical area; - defining ( 102 ) for each risk category a plurality of keywords semantically related to the risk category and storing said plurality of keywords in a data structure; - for each risk category, extracting ( 103 ) first risk data by consulting informative websites and social media and carrying out a computer sentiment analysis of first textual contents published in, or associated with, a first relatively long time period on said informative websites and social media, where the first textual contents concern and / or mention said geographical area and are selected using the keywords semantically related to the risk category; - determining ( 104 ) a basic risk index associated with the geographical area starting from the first risk data; - receiving or generating a ( 106) request to determine a. n instantaneous risk index associated with the geographical area; - for each risk category, extracting ( 107 ) second risk data by consulting informative websites and social media and carrying out a computer sentiment analysis of second textual contents published in, or associated with, a second relatively short time period on said informative websites and social media, where the second textual contents concern and / or mention said geographical area and are selected using the keywords semantically related to the risk category; - determining ( 108 ) an instantaneous risk index associated with the geographical area starting from the second risk data; - comparing ( 109 ) the instantaneous risk index and the basic risk index; generate and send ( 110 ) a notification and / or control signal if from the comparing step it is found that the instantaneous risk index is higher than the basic risk index

The invention provides a method and device for evaluating the risk of a flight plan, and relates to the technical field of air traffic control automation. The method for evaluating the risk of the flight plan comprises the following steps: determining a management side influence factor and a service side influence factor based on a plurality of preset risk categories; constructing a management side risk assessment model and a service side risk assessment model according to a risk value calculation formula of the management side influence factors and the service side influence factors; selecting a target risk assessment model from the management side risk assessment model and the service side risk assessment model according to the flight plan to be assessed and the assessment purpose thereof; based on the target risk assessment model, calculating a risk value of the management side influence factor or the service side influence factor according to the flight plan to be assessed; and according to the risk value of the management side influence factor or the service side influence factor, carrying out weighted calculation to obtain a comprehensive risk value as an evaluation result. The risk value can be automatically calculated, the efficiency is high, and the credibility is high.

This invention discloses a system risk monitoring method, device, computer equipment, and medium. It constructs a risk characteristic database covering the entire lifecycle of ship equipment development, collects data from existing ship projects, defines six primary risks and their subordinate secondary risks, and sets scoring criteria to map the risk characteristics of existing projects into vectors and store them in association. It extracts information from projects to be reviewed, calculates the dynamic risk index of each secondary risk category of the project using a dynamic risk index model, and classifies the risk level. It establishes target risk characteristic vectors based on the ship type and development stage of the project to be reviewed. It calculates the cosine similarity between the target risk characteristic vector and the risk characteristic vectors of all existing ship projects in the full lifecycle risk characteristic database, clusters similar project groups, statistically analyzes risk data to generate prompt statements, and establishes a risk response plan library to select an initial plan and determine the execution plan.

The invention provides a risk assessment method and device, equipment, a storage medium and a program product, and relates to the field of artificial intelligence. The method comprises the following steps: constructing an object relation graph according to feature data of an object; wherein the object entities are used as nodes of the object relation graph, and edges of the object relation graph are established based on feature similarity among the nodes; processing the object relation graph based on a graph neural network model to obtain aggregation node features; the graph neural network model is used for traversing the object relation graph based on a Pearson's correlation coefficient between node features and aggregating neighbor node features subjected to traversal sampling; and according to the aggregation node features, performing classification processing through a classifier to obtain a risk assessment result of the object, the risk assessment result comprising a risk category. According to the method, the relevance risk between enterprises can be found, and mass enterprise data can be efficiently and accurately processed.

The specification relates to the technical field of artificial intelligence, and particularly discloses a target object risk identification method and device, wherein the method comprises: receiving a risk detection request; in response to the risk detection request, obtaining target object data corresponding to a target object identifier; the target object data comprises first type attribute feature data and second type attribute feature data; in the case where there is missing index data and / or erroneous index data in the second type attribute feature data, using a naive Bayes decision algorithm to perform data completion on the missing index data and / or correction on the erroneous index data in the second type attribute feature data, to obtain the second type attribute feature data after completion and / or correction; and performing clustering analysis on the first type attribute feature data and the second type attribute feature data after completion and / or correction, to obtain a risk category corresponding to the target object identifier. The above scheme can improve the accuracy of target object risk identification and adapt to changes in the business environment.

This application discloses a method, apparatus, and system for rapid response and automatic review of claims, addressing the problem that computer systems cannot automatically and rapidly respond to claims events triggering the generation of payout calculation results. The method includes: receiving a claims case identifier; acquiring its multi-dimensional risk characteristic data; determining whether the case belongs to a low-risk category that can be automatically processed through a risk stratification model based on a configurable rule set; if so, then executing intelligent claims processing and integrated risk control in parallel, i.e., determining the payout amount based on policy and claim information matching rules, while simultaneously comparing the case data with a dynamically configured risk rule set in real time; if the risk control passes, automatically outputting a review instruction. This invention achieves full-process automation of life insurance claims processing from risk stratification and intelligent claims processing to real-time risk control, significantly improving processing efficiency and accuracy while strictly ensuring risk and compliance.

The invention discloses an automatic compliance review workflow generation method based on large model agent arrangement, which comprises the following steps: extracting a text feature vector of an input document through a document classification model, and obtaining a key term list based on the text feature vector; semantic intention analysis is carried out on the key clause list, the clause type of each clause is determined, and potential risk points are identified; if the clause type is a high-risk type, matching law and regulation clauses in a knowledge base through a rule engine, obtaining a comparison result, identifying non-compliance items, and evaluating the number and severity of the non-compliance items; when the number of non-compliance items exceeds a preset threshold value, a sequence generation model is adopted to generate a dynamic step sequence based on the review context, and input data of each step in the dynamic step sequence are obtained; according to the method, full-chain automatic processing from risk identification to compliance evaluation is realized, the examination efficiency and accuracy are remarkably improved, and efficient technical support is provided for compliance management of complex documents.